A Big Data-Based Method and System for Ecological and Environmental Pollution Risk Assessment

Abstract

This invention discloses a method and system for ecological and environmental pollution risk assessment based on big data, belonging to the field of environmental pollution assessment technology. This invention uses rainstorm events as triggers to construct an event-aligned dataset, integrates rainfall radar, remote sensing cover, soil moisture, topographic runoff, agricultural activity agents, and river and lake water quality and flow-related data, inverts the potential release index of nutrients in sub-basins and generates event propagation weights, and combines an event-driven coupled assessment model constructed with historical events to output transient inflow load and composite risk indicators, further providing risk contribution ranking and minimum intervention set, realizing integrated early warning and decision support for pre-rainstorm source control, in-rainstorm interception, and post-rainstorm restoration.

Description

Technical Field

[0001] This invention relates to the field of environmental pollution assessment technology, and in particular to a method and system for assessing ecological and environmental pollution risks based on big data. Background Technology

[0002] In watersheds with a high proportion of agricultural non-point source pollution and downstream lakes, reservoirs, or slow-flowing river sections, heavy rainfall events such as torrential rains or typhoons can trigger slope runoff and ditch erosion in a short period of time, causing nutrients such as nitrogen and phosphorus to rapidly enter the river in a pulse-like manner and be transported downstream to receiving water bodies. Receiving water bodies are susceptible to secondary environmental risks after heavy rainfall, including algal blooms, a sharp drop in dissolved oxygen, and the resurgence of black and odorous water, due to factors such as water temperature stratification, retention time, and endogenous release from bottom sediments.

[0003] Currently, Chinese invention patent application number 202311736253.0 discloses a method and system for analyzing ecological and environmental pollution risks in watersheds of ecologically sensitive areas. The method includes the following steps: constructing a knowledge graph of new pollutants; constructing several historical data sets for measuring new pollutants and several historical data sets for the impact of ecological and environmental pollution; constructing a training sample set for historical ecological and environmental pollution risk analysis; constructing an ecological and environmental pollution risk analysis model; collecting real-time water quality data of the current ecologically sensitive watershed and performing data filtering; and conducting ecological and environmental pollution risk analysis to obtain the results of the current ecological and environmental pollution risk analysis of the watershed. The system includes a knowledge graph construction unit, an indicator data construction unit, a training sample construction unit, a model construction unit, a data filtering unit, and a pollution risk analysis unit connected sequentially.

[0004] The aforementioned technologies are insufficient to predict hourly-level compound pollution risks of non-point source nutrients under strong event conditions based on multi-source big data and simultaneously output explainable attribution and minimal intervention strategies. Summary of the Invention

[0005] The technical problem solved by this invention is that existing technologies are unable to predict hourly compound pollution risks of non-point source nutrients based on multi-source big data under strong event conditions, and simultaneously output explainable attribution and minimal intervention strategies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A big data-based method for assessing ecological and environmental pollution risks includes the following steps:

[0008] Step S1: Collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity agent data, river and lake water quality data, and hydrological flow data; clean, synchronize in time, and spatially grid the data to form raw event perception data and construct an event-aligned dataset.

[0009] Step S2: Based on the event-aligned dataset, invert the potential release index set of nutrients in the sub-basin;

[0010] Step S3: Construct an event propagation graph, calculate edge weights based on event alignment data, and output the event propagation weight set;

[0011] Step S4: Construct an event-driven coupled assessment model. Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled assessment model, and output the transient inflow load prediction set and the composite pollution risk index set.

[0012] Step S5: Generate risk zoning early warning results and risk contribution ranking set based on the composite pollution risk index set, and select the corresponding minimum intervention set to output to the control end.

[0013] Preferably, step S1 includes the following sub-steps:

[0014] Step S101: Collect historical rainstorm event data, rainfall radar data, and weather forecast data, and merge the rainfall radar data and weather forecast data into raw rainfall event data;

[0015] Step S102: Collect remote sensing cover data, soil moisture data, and topographic runoff data to form raw runoff-sensitive data;

[0016] Step S103: Collect agricultural activity proxy data to form raw human activity proxy data;

[0017] Step S104: Collect river and lake water quality data and hydrological flow data to form raw receiving response data;

[0018] Historical rainstorm event data, raw rainfall event data, raw runoff-sensitive data, raw human activity proxy data, and raw reception response data are merged and output as raw event perception data and stored in the database.

[0019] Preferably, the event alignment dataset is constructed as follows:

[0020] Identify the start and end times of rainstorm events based on raw rainfall event data;

[0021] The raw event perception data is sliced ​​into event windows according to the start and end times of the rainstorm event. The event windows include a pre-rainstorm window, a rainstorm window, and a post-rainstorm window.

[0022] Perform temporal interpolation alignment, spatial raster resampling alignment, and outlier removal on the event window slice data, and output the event-aligned dataset.

[0023] Preferably, step S2 includes the following sub-steps:

[0024] Step S201: Extract cover type, crop growth status, soil moisture content index and slope runoff index based on the event alignment dataset to form a non-point source sensitive feature set;

[0025] Step S202: Construct a nutrient release prior parameter set based on the surface source sensitive feature set and the original human activity proxy data;

[0026] Step S203: Perform regression fitting between the prior parameter set of nutrient release and the event-aligned dataset to output the potential release index set of nutrients in the sub-basin.

[0027] Preferably, the original human activity proxy data includes agricultural input transaction record data, agricultural machinery operation trajectory data, and application time series estimation data, and the nutrient release prior parameter set is constructed as follows:

[0028] Identify suspected application periods based on agricultural input transaction records and agricultural machinery operation trajectory data;

[0029] By associating suspected application periods with cover type and crop growth status, an estimated application intensity can be obtained.

[0030] Based on the estimated application intensity, output the application time series estimation data and incorporate it into the nutrient release prior parameter set.

[0031] Preferably, step S3 includes the following sub-steps:

[0032] Step S301: Using sub-basins as nodes, construct an event propagation graph with the land parcel confluence paths, ditch connectivity, tributary confluence relationships, and pipeline overflow paths as edges;

[0033] Among them, when the overflow monitoring quantity or overflow status corresponding to the overflow path of the pipeline meets the preset overflow triggering conditions, the corresponding edge is marked as an active edge; otherwise, it is marked as an inactive edge.

[0034] Step S302: Calculate the flow generation driving factor and transport retardation factor for each side based on the event alignment dataset;

[0035] Step S303: Normalize and couple the runoff generation driving factor and the transport hindrance factor and map them as time-varying edge weights, update them according to the pre-rainstorm window, during-rainstorm window and post-rainstorm window, and output the event propagation weight set.

[0036] Preferably, step S4 includes the following sub-steps:

[0037] Step S401: Construct a historical training sample set based on the historical event alignment dataset corresponding to multiple historical rainstorm events;

[0038] Step S402: The historical potential release index of nutrients and the historical event propagation weight generated in steps S2 and S3 are used as training inputs. The historical inflow load reference sequence formed by historical river and lake water quality data and historical hydrological flow data, as well as the algal bloom trigger reference marker, hypoxia risk reference marker and secondary black and odor risk reference marker constructed based on the historical sensitivity parameter set of rivers and lakes, are used as supervision information to fit the event-driven coupling evaluation model.

[0039] When the deviation between the composite pollution risk index output by the event-driven coupled assessment model and the event response of historical river and lake water quality data exceeds the preset consistency threshold, the source strength mapping or path propagation related parameters are updated and intra-event calibration is completed.

[0040] Step S403: Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled evaluation model, and output the transient inflow load prediction set;

[0041] Step S404: Based on the transient inflow load prediction set and river and lake water quality data and hydrological flow data, calculate the probability of algal bloom triggering, the probability of hypoxia risk and the probability of secondary black and odorous pollution risk to form a composite pollution risk index set.

[0042] Preferably, step S5 includes the following sub-steps:

[0043] Step S501: Generate risk zoning early warning results based on the composite pollution risk index set;

[0044] Step S502, decompose the contribution of the composite pollution risk index set based on the event propagation graph, specifically as follows:

[0045] While keeping the event propagation weight set unchanged, a preset amplitude perturbation is applied to the potential release quantity index corresponding to each source node.

[0046] Calculate the risk increments of algal bloom trigger probability, hypoxia risk probability, and secondary black and odorous environment risk probability before and after the disturbance;

[0047] Define the risk increment as the contribution of the corresponding source strength node and sort them, then output the risk contribution sort set;

[0048] Step S503: Based on the risk contribution ranking set and the preset resource constraint rules, the minimum intervention set is calculated.

[0049] Preferably, the selection logic for the minimum intervention set is as follows:

[0050] When the probability of algal bloom triggering or the probability of hypoxia risk are higher than the preset threshold, the interception measures corresponding to the source strength nodes ranked higher in the risk contribution ranking set are selected first, and the source control timing is matched with the window before the rainstorm.

[0051] When the probability of secondary black and odorous pollution is higher than the preset threshold, oxygenation or scheduling measures corresponding to the receiving nodes with the highest risk contribution ranking will be selected first, and the repair sequence of the window after the rainstorm will be matched.

[0052] The source control timing sequence and the repair timing sequence are combined and output as a minimum intervention set and input to the control terminal.

[0053] An ecological and environmental pollution risk assessment system based on big data includes a data acquisition module, a data inversion module, a propagation analysis module, a driving coupling module, and a risk determination module;

[0054] The data acquisition module is used to collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity proxy data, river and lake water quality data, and hydrological flow data. It then performs cleaning, time synchronization, and spatial gridding to form raw event perception data and construct an event-aligned dataset.

[0055] The data inversion module is used to invert the potential release index set of nutrients in a sub-basin based on the event-aligned dataset.

[0056] The propagation analysis module is used to construct an event propagation graph, calculate edge weights based on event alignment data, and output an event propagation weight set.

[0057] The driving coupling module is used to input the potential release index set, the event propagation weight set, and the event alignment dataset into the event-driven coupling evaluation model, and output the transient inflow load prediction set and the composite pollution risk index set.

[0058] The risk assessment module is used to generate risk zoning early warning results and risk contribution ranking sets based on the composite pollution risk indicator set, and select the corresponding minimum intervention set to output to the control end.

[0059] The beneficial effects of this invention are as follows: This invention uses rainstorm events as triggers to construct an event-aligned dataset, integrates rainfall radar, remote sensing cover, soil moisture, topographic runoff, agricultural activity agents, and river and lake water quality and flow-related data, inverts the potential release index of nutrients in sub-basins and generates event propagation weights, combines an event-driven coupled assessment model constructed from historical events, outputs transient inflow load and composite risk indicators, and further provides a risk contribution ranking and minimum intervention set, realizing integrated early warning and decision support for pre-rainstorm source control, in-rainstorm interception, and post-rainstorm restoration. Attached Figure Description

[0060] Figure 1 A flowchart illustrating the steps of a big data-based ecological environment pollution risk assessment method provided in one embodiment of the present invention;

[0061] Figure 2This is a basic flowchart of an ecological environment pollution risk assessment system based on big data, provided as an embodiment of the present invention. Detailed Implementation

[0062] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0063] Example 1, referring to Figure 1 This paper provides a big data-based method for assessing ecological and environmental pollution risks, which includes the following steps:

[0064] Step S1: Collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity agent data, river and lake water quality data, and hydrological flow data. Clean, synchronize and spatially grid the data to form raw event perception data and construct an event-aligned dataset.

[0065] Step S2: Based on the event-aligned dataset, invert the potential release index set of nutrients in the sub-basin.

[0066] Step S3: Construct an event propagation graph, calculate edge weights based on event alignment data, and output the event propagation weight set.

[0067] Step S4: Construct an event-driven coupled assessment model. Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled assessment model, and output the transient inflow load prediction set and the composite pollution risk index set.

[0068] Step S5: Generate risk zoning early warning results and risk contribution ranking set based on the composite pollution risk index set, and select the corresponding minimum intervention set to output to the control end.

[0069] This invention uses rainstorm events as triggers to construct an event-aligned dataset, integrating rainfall radar, remote sensing cover, soil moisture, topographic runoff, agricultural activity agents, and river and lake water quality and flow-related data. It inverts the potential release index of nutrients in sub-basins and generates event propagation weights. Combined with an event-driven coupled assessment model constructed from historical events, it outputs transient inflow loads and composite risk indicators, and further provides a risk contribution ranking and minimum intervention set, achieving integrated early warning and decision support for pre-rainstorm source control, in-rainstorm interception, and post-rainstorm restoration.

[0070] This embodiment takes a watershed with significant upstream agricultural non-point source pollution and downstream lakes and bays prone to algal blooms as the application object. It conducts an event-driven ecological and environmental pollution risk assessment for a heavy rainfall event and uses the assessment results to guide the coordinated management and control of source control before the rainstorm, interception during the rainstorm, and emergency restoration after the rainstorm. Taking the rainstorm event as the trigger condition, it integrates rainfall, runoff sensitivity, human activity proxy, and receiving response data around the complex propagation chain, and outputs a set of complex pollution risk indicators, risk zoning early warning results, risk contribution ranking set, and minimum intervention set for sub-basins, river sections, and lake bays.

[0071] Step S1 includes the following sub-steps:

[0072] Step S101: Collect historical rainstorm event data, rainfall radar data, and weather forecast data, and merge the rainfall radar data and weather forecast data into raw rainfall event data.

[0073] Step S102: Collect remote sensing cover data, soil moisture data, and topographic runoff data to form raw runoff sensitive data.

[0074] Step S103: Collect agricultural activity proxy data to form raw human activity proxy data.

[0075] Step S104: Collect river and lake water quality data and hydrological flow data to form raw receiving response data.

[0076] Historical rainstorm event data, raw rainfall event data, raw runoff-sensitive data, raw human activity proxy data, and raw reception response data are merged and output as raw event perception data and stored in the database.

[0077] The raw rainfall event data, raw runoff-sensitive data, raw human activity proxy data, and raw reception response data are merged and output as raw event perception data.

[0078] First, the data acquisition and raw event perception data construction process is executed. Rainfall radar data and weather forecast data are acquired from meteorological radar and short-term forecast services to form raw rainfall event data. Simultaneously, remote sensing cover data, soil moisture data, and topographic runoff data are acquired from remote sensing products and geographic information databases to form raw runoff-sensitive data. Next, agricultural activity proxy data is acquired from external or industry data interfaces related to agricultural production to form raw human activity proxy data. Finally, river and lake water quality data and hydrological flow data are acquired from online monitoring stations of rivers and reservoirs, historical monitoring databases, and hydrological stations to form raw receiving response data. Subsequently, the raw rainfall event data, raw runoff-sensitive data, raw human activity proxy data, and raw receiving response data are aggregated according to a unified spatiotemporal index to obtain the raw event perception data. This stage ensures that the data input for rainfall triggering, non-point source release, transport paths, and receiving responses in subsequent steps is consistent and traceable.

[0079] The event alignment dataset is constructed as follows:

[0080] Identify the start and end times of rainstorm events based on raw rainfall event data.

[0081] The raw event perception data is sliced ​​into event windows according to the start and end times of the rainstorm event. The event windows include the pre-rainstorm window, the rainstorm window, and the post-rainstorm window.

[0082] Perform temporal interpolation alignment, spatial raster resampling alignment, and outlier removal on the event window slice data, and output the event-aligned dataset.

[0083] After obtaining the raw event perception data, the start and end times of the rainstorm event are identified based on the raw rainfall event data. Specifically, when the raster rainfall intensity in the target area meets the intensity threshold condition for a continuous preset duration, or when the cumulative rainfall meets the cumulative threshold condition, it is marked as the start of the rainstorm event. When the rainfall intensity remains below the threshold for a preset duration, it is marked as the end of the rainstorm event. Then, the raw event perception data is sliced ​​into event windows using the start and end times of the rainstorm event as anchor points to generate pre-rainstorm window data, mid-rainstorm window data, and post-rainstorm window data. Temporal interpolation alignment, spatial raster alignment, outlier removal, and missing data marking are performed on the event window slice data to finally form an event-aligned dataset. Through this alignment method, multi-source data with inconsistent resolution and sampling frequency are unified into a computable object serving the same rainstorm event, so that subsequent source strength inversion, propagation weight calculation, and composite risk assessment have a consistent input basis.

[0084] Step S2 includes the following sub-steps:

[0085] Step S201: Extract cover type, crop growth status, soil moisture content index and slope runoff index based on the event alignment dataset to form a non-point source sensitive feature set.

[0086] Step S202: Construct a prior parameter set for nutrient release based on the surface source sensitive feature set and the original human activity proxy data.

[0087] Step S203: Perform regression fitting between the prior parameter set of nutrient release and the event-aligned dataset to output the potential release index set of nutrients in the sub-basin.

[0088] Based on the event-aligned dataset, the inversion of the potential release index set of nutrients in the sub-basin is performed. First, cover type, crop growth status, soil moisture content before the rainstorm, and slope runoff index are extracted from the event-aligned dataset to form a non-source heat pump sensitive feature set. The possible application time and intensity are estimated using the original human activity proxy data to construct a nutrient release prior parameter set. This prior parameter set is then associated with the non-source heat pump sensitive feature set. In practice, agricultural input transaction record data and agricultural machinery operation trajectory data are coupled with crop growth status in a temporal sequence to identify suspected application windows. Spatially, suspected application windows are matched to the corresponding sub-basin or plot grid range, and application intensity estimates and application time sequence estimates are output. Subsequently, the application intensity estimates, application time sequence estimates, and soil moisture content before the rainstorm together constitute the nutrient release prior parameter set. Weakly supervised fitting is then performed by combining the response characteristics of upstream inflow water quality and flow in historical rainstorm events to output the potential release index set of nutrients in the sub-basin. The sub-basin nutrient potential release index set is used to describe the potential scale of nutrients in each sub-basin that can be washed away and enter the river network under the conditions of this rainstorm event, providing a quantitative basis for the source strength assignment on the subsequent propagation map.

[0089] The original human activity proxy data includes agricultural input transaction records, agricultural machinery operation trajectory data, and application time series estimation data. The construction of the nutrient release prior parameter set is as follows:

[0090] Suspected application periods were identified based on agricultural input transaction records and agricultural machinery operation trajectory data.

[0091] By associating the suspected application period with the cover type and crop growth status, an estimated application intensity can be obtained.

[0092] Based on the estimated application intensity, output the application time series estimation data and incorporate it into the nutrient release prior parameter set.

[0093] In this embodiment, the agricultural activity agency data includes at least agricultural input transaction record data and agricultural machinery operation trajectory data. The purchase time or delivery time indicated by the agricultural input transaction record data is used as the application possibility time prior, and the operation location range indicated by the agricultural machinery operation trajectory data is used as the application possibility space prior. The two are matched with the centralized cover type and crop growth status of the event alignment dataset to generate application intensity estimates and application time series estimates.

[0094] Step S3 includes the following sub-steps:

[0095] Step S301: Using sub-basins as nodes, construct an event propagation graph with the land parcel confluence paths, ditch connectivity, tributary confluence relationships, and pipeline overflow paths as edges.

[0096] Specifically, when the overflow monitoring quantity or overflow status corresponding to the overflow path of the pipeline meets the preset overflow triggering conditions, the corresponding edge is marked as an active edge; otherwise, it is marked as an inactive edge.

[0097] Step S302: Calculate the flow generation driving factor and transport hindrance factor for each side based on the event alignment dataset.

[0098] Step S303: Map the flow generation driving factor and the transport obstruction factor to edge weights and output the event propagation weight set.

[0099] Step S4 involves constructing an event-driven coupled evaluation model based on a historical event alignment dataset, specifically including:

[0100] The database retrieves historical event alignment datasets corresponding to multiple historical rainstorm events. The pre-rainstorm window data, rainstorm window data, and post-rainstorm window data of each historical rainstorm event are matched according to the spatial index of sub-basin, river segment, and lake bay to form a historical event sample set. Each sample in the historical event sample set contains at least an alignment sequence of rainfall radar data, meteorological forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity proxy data, river and lake water quality data, and hydrological flow data within the same event window.

[0101] For each historical event sample, based on the corresponding historical event alignment dataset, according to the processing logic of steps S2 and S3, the historical nutrient potential release index and historical event propagation weight are generated respectively, and the historical nutrient potential release index and historical event propagation weight are bound to the corresponding historical event alignment dataset to form a historical training input set.

[0102] For each historical event sample, within the window during and after the rainstorm, a reference sequence of historical inflow loads corresponding to sub-basins or inflow sections is constructed using the combined response of historical river and lake water quality data and historical hydrological flow data as a reference. This historical inflow load reference sequence is then used as training and supervision information for transient inflow loads.

[0103] For each historical event sample, the response changes of historical river and lake water quality data within the event window of the bay or receiving water body are used as a reference. Combined with the historical sensitivity parameter set of rivers and lakes, algal bloom trigger reference markers, hypoxia risk reference markers, and secondary black and odorous risk reference markers are constructed. The above reference markers are used as training and supervision information for composite pollution risk indicators.

[0104] Using historical training input sets as input objects and historical river load reference sequences and composite pollution risk reference markers as supervisory information, a parameter mapping relationship is established and fitted to obtain an event-driven coupled assessment model. The model is calibrated based on the intra-event bias of different historical rainstorm events to form a model parameter set for the assessment of this rainstorm event.

[0105] Historical rainstorm events are divided into a model building event subset and a model validation event subset according to a preset ratio. When the consistency between the composite pollution risk index of the model validation event subset and the event response of historical river and lake water quality data meets the preset consistency rule, the event-driven coupled evaluation model is determined to be a valid model. Otherwise, the mapping parameters of the agricultural activity proxy data weights or event propagation weights in the historical training input set are updated backtrackingly until the preset consistency rule is met.

[0106] Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled evaluation model, and output the transient inflow load prediction set.

[0107] Based on the transient inflow load prediction set and river and lake water quality data and hydrological flow data, the probability of algal bloom triggering, the probability of hypoxia risk and the probability of secondary black and odorous pollution risk are calculated to form a composite pollution risk index set.

[0108] The event-driven coupled evaluation model is constructed as follows:

[0109] Based on the historical event alignment dataset corresponding to multiple historical rainstorm events, the historical nutrient potential release index and historical event propagation weight are generated according to steps S2 and S3 as training inputs.

[0110] Using historical river and lake water quality data and historical hydrological flow data to form a historical inflow load reference sequence, as well as algal bloom triggering reference markers, hypoxia risk reference markers, and secondary black and odorous risk reference markers constructed based on historical river and lake sensitivity parameter sets as supervisory information, an event-driven coupled assessment model was fitted and intra-event calibration was completed.

[0111] The logic of the event-driven coupled evaluation model is as follows:

[0112] The potential release index set is used as the source strength input, the event propagation weight set is used as the output input, and river and lake water quality data and hydrological flow data are used as the receiving verification input.

[0113] The transient inflow load prediction set was calculated.

[0114] Based on the transient inflow load prediction set and the historical sensitivity parameter set of rivers and lakes, the probability of algal bloom triggering, the probability of hypoxia risk and the probability of secondary black and odorous pollution risk are calculated to form a composite pollution risk index set.

[0115] The event-driven coupled assessment model inputs the potential release index set, the event propagation weight set, and the event alignment dataset to output a transient inflow load prediction set and a composite pollution risk index set. The model uses the potential release index set as the source strength input, the event propagation weight set as the transport input, and river and lake water quality data and hydrological flow data as receiving state constraints and intra-event verification inputs. It also uses historical sensitivity parameter sets of rivers and lakes to correct for vulnerability at the receiving end. The model first calculates the transient inflow load prediction set and the composite pollution risk index set for each sub-basin within the storm window based on the source strength and transport weights. The inflow load prediction set is combined with real-time or event-specific hydrological flow data to form a time-series prediction of nutrient flux in the river section. For lake bay nodes, the event-driven coupled assessment model couples the time-series prediction of nutrient flux with the historical sensitivity parameter set of rivers and lakes, outputting the probability of algal bloom triggering, the probability of hypoxia risk, and the probability of secondary black and odorous pollution risk, thus forming a composite pollution risk index set. If there is a deviation between the online water quality data and the output of the event-driven coupled assessment model during the event, the event-driven coupled assessment model updates the credibility label of source strength and transport weight according to the direction of deviation, realizing dynamic correction within the event.

[0116] Step S5 includes the following sub-steps:

[0117] Step S501: Generate risk zoning early warning results based on the composite pollution risk index set.

[0118] Step S502, decompose the contribution of the composite pollution risk index set based on the event propagation graph, specifically as follows:

[0119] While keeping the event propagation weight set unchanged, a preset amplitude perturbation is applied to the potential release quantity index corresponding to each source strong node.

[0120] Calculate the risk increments of algal bloom trigger probability, hypoxia risk probability, and secondary black and odorous environment risk probability before and after the disturbance.

[0121] The risk increment is defined as the contribution of the corresponding source node and sorted, and the risk contribution sorted set is output.

[0122] Step S503: Based on the risk contribution ranking set and the preset resource constraint rules, the minimum intervention set is calculated.

[0123] The selection logic for the minimum intervention set is as follows:

[0124] When the probability of algal bloom triggering or the probability of hypoxia risk is higher than the preset threshold, the interception measures corresponding to the source strength nodes ranked higher in the risk contribution ranking set are selected first, and the source control timing is matched with the window before the rainstorm.

[0125] When the probability of secondary black and odorous pollution is higher than the preset threshold, oxygenation or scheduling measures corresponding to the receiving nodes that rank higher in the risk contribution ranking will be selected first, and the repair sequence of the window after the rainstorm will be matched.

[0126] The source control timing sequence and the repair timing sequence are combined and output as a minimum intervention set and input to the control terminal.

[0127] After obtaining the set of composite pollution risk indicators, the following steps are performed: early warning generation, contribution decomposition, and minimum intervention set back-calculation. First, risk zoning early warning results are generated at three levels—sub-basin, river segment, and lake bay—based on the composite pollution risk indicator set, clarifying the risk level and possible trigger time windows. Then, the contribution of the composite pollution risk indicator set is decomposed based on the event propagation graph, outputting a risk contribution ranking set.

[0128] In practice, the composite risk probability of the lake bay node is taken as the target output, and the marginal perturbation calculation is performed on the source strength node in the propagation diagram: under the condition of keeping the event propagation weight set unchanged, the potential release index of a single sub-basin is perturbed by amplitude, the increase of the risk probability of algal bloom, hypoxia and black and odorous before and after the perturbation is compared, the risk probability increase is defined as the contribution of the corresponding source strength node, and the contribution is sorted to form a risk contribution ranking set.

[0129] Based on the risk contribution ranking set and the preset resource constraint rules, the minimum intervention set is calculated: when the probability of algal bloom triggering or hypoxia risk exceeds the preset threshold, the source control or interception measures corresponding to the source strength nodes ranked higher are selected first and matched with the pre-rainstorm window or the rainstorm window.

[0130] When the probability of secondary black and odorous pollution exceeds a preset threshold, priority is given to selecting oxygenation, scheduling, or emergency remediation measures corresponding to the top-ranked receiving nodes, and matching them with the post-rainstorm window. The above source control time series and remediation time series are combined and output as a minimum intervention set and input into the management and control terminal to guide collaborative governance under this rainstorm event.

[0131] Example 2, refer to Figure 2 This paper presents a big data-based ecological and environmental pollution risk assessment system, which includes a data acquisition module, a data inversion module, a propagation analysis module, a driving coupling module, and a risk determination module.

[0132] The data acquisition module is used to collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity agent data, river and lake water quality data, and hydrological flow data. It performs cleaning, time synchronization, and spatial gridding to form raw event perception data and constructs an event-aligned dataset.

[0133] The data inversion module is used to invert the potential release index set of nutrients in a sub-basin based on the event-aligned dataset.

[0134] The propagation analysis module is used to construct an event propagation graph, calculate edge weights based on event alignment data, and output an event propagation weight set.

[0135] The driving coupling module is used to input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupling evaluation model, and output the transient inflow load prediction set and the composite pollution risk index set.

[0136] The risk assessment module is used to generate risk zoning early warning results and risk contribution ranking sets based on the composite pollution risk indicator set, and select the corresponding minimum intervention set to output to the control end.

[0137] This invention uses rainstorm events as triggers to construct an event-aligned dataset, which integrates rainfall radar, remote sensing cover, soil moisture, topographic runoff, agricultural activity agents, and river and lake water quality and hydrological flow related data. It can achieve coordinated characterization of surface source release conditions, transport connectivity characteristics, and receiving response status within the same event window.

[0138] By inverting the potential release index set of nutrients in sub-basins and constructing an event propagation map, combined with an event-driven coupled assessment model constructed from historical event-aligned datasets, this method can output a transient inflow load prediction set and a composite pollution risk index set of algal blooms, hypoxia, and secondary black and odorous pollution, enabling zone-level risk identification and early warning for key periods before, during, and after rainstorms.

[0139] Furthermore, this invention decomposes risk contribution based on a propagation graph, forms a risk contribution ranking set, and reverse-engineers the minimum intervention set under resource constraints. This can clearly identify the upstream sub-basins for priority source control, the key channels for priority interception, the downstream lake bays for priority restoration, and the corresponding time windows. Thus, the risk assessment results are directly transformed into a feasible collaborative governance strategy for rainstorm events, significantly improving the basin's response efficiency and management sophistication in the face of sudden non-point source pollution and complex secondary risks.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0141] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for assessing ecological and environmental pollution risks based on big data, characterized in that, Includes the following steps: Step S1: Collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity agent data, river and lake water quality data, and hydrological flow data; clean, synchronize in time, and spatially grid the data to form raw event perception data; and construct an event-aligned dataset. Step S2: Based on the event-aligned dataset, invert the potential release index set of nutrients in the sub-basin; Step S3: Construct an event propagation graph, calculate edge weights based on event alignment data, and output the event propagation weight set; Step S4: Construct an event-driven coupled assessment model. Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled assessment model, and output the transient inflow load prediction set and the composite pollution risk index set. Step S5: Generate risk zoning early warning results and risk contribution ranking set based on the composite pollution risk index set, and select the corresponding minimum intervention set to output to the control end; The event alignment dataset is constructed as follows: Identify the start and end times of rainstorm events based on raw rainfall event data; The raw event perception data is sliced ​​into event windows according to the start and end times of the rainstorm event. The event windows include a pre-rainstorm window, a rainstorm window, and a post-rainstorm window. Perform temporal interpolation alignment, spatial raster resampling alignment, and outlier removal on the event window slice data, and output the event-aligned dataset; Step S2 includes the following sub-steps: Step S201: Extract cover type, crop growth status, soil moisture content index and slope runoff index based on the event alignment dataset to form a non-point source sensitive feature set; Step S202: Construct a nutrient release prior parameter set based on the surface source sensitive feature set and the original human activity proxy data; Step S203: Perform regression fitting between the prior parameter set of nutrient release and the event-aligned dataset to output the potential release index set of nutrients in the sub-basin. The original human activity proxy data includes agricultural input transaction records, agricultural machinery operation trajectory data, and application time series estimation data. The construction of the nutrient release prior parameter set is as follows: Identify suspected application periods based on agricultural input transaction records and agricultural machinery operation trajectory data; By associating suspected application periods with cover type and crop growth status, an estimated application intensity can be obtained. Based on the estimated application intensity, output the application time series estimation data and incorporate it into the nutrient release prior parameter set; Step S3 includes the following sub-steps: Step S301: Using sub-basins as nodes, construct an event propagation graph with the land parcel confluence paths, ditch connectivity, tributary confluence relationships, and pipeline overflow paths as edges; Among them, when the overflow monitoring quantity or overflow status corresponding to the overflow path of the pipeline meets the preset overflow triggering conditions, the corresponding edge is marked as an active edge; otherwise, it is marked as an inactive edge. Step S302: Calculate the flow generation driving factor and transport retardation factor for each side based on the event alignment dataset; Step S303: Normalize and couple the runoff generation driving factor and the transport hindrance factor and map them as time-varying edge weights, update them according to the pre-rainstorm window, during-rainstorm window and post-rainstorm window, and output the event propagation weight set.

2. The method for assessing ecological and environmental pollution risks based on big data as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S101: Collect historical rainstorm event data, rainfall radar data, and weather forecast data, and merge the rainfall radar data and weather forecast data into raw rainfall event data; Step S102: Collect remote sensing cover data, soil moisture data, and topographic runoff data to form raw runoff-sensitive data; Step S103: Collect agricultural activity proxy data to form raw human activity proxy data; Step S104: Collect river and lake water quality data and hydrological flow data to form raw receiving response data; Historical rainstorm event data, raw rainfall event data, raw runoff-sensitive data, raw human activity proxy data, and raw reception response data are merged and output as raw event perception data and stored in the database.

3. The method for assessing ecological and environmental pollution risks based on big data as described in claim 1, characterized in that, Step S4 includes the following sub-steps: Step S401: Construct a historical training sample set based on the historical event alignment dataset corresponding to multiple historical rainstorm events; Step S402: The historical potential release index of nutrients and the historical event propagation weight generated in steps S2 and S3 are used as training inputs. The historical inflow load reference sequence formed by historical river and lake water quality data and historical hydrological flow data, as well as the algal bloom trigger reference marker, hypoxia risk reference marker and secondary black and odor risk reference marker constructed based on the historical sensitivity parameter set of rivers and lakes, are used as supervision information to fit the event-driven coupling evaluation model. When the deviation between the composite pollution risk index output by the event-driven coupled assessment model and the event response of historical river and lake water quality data exceeds the preset consistency threshold, the source strength mapping or path propagation related parameters are updated and intra-event calibration is completed. Step S403: Input the potential release index set, event propagation weight set, and event alignment dataset into the event-driven coupled evaluation model, and output the transient inflow load prediction set; Step S404: Based on the transient inflow load prediction set and river and lake water quality data and hydrological flow data, calculate the probability of algal bloom triggering, the probability of hypoxia risk and the probability of secondary black and odorous pollution risk to form a composite pollution risk index set.

4. The method for assessing ecological and environmental pollution risks based on big data as described in claim 3, characterized in that, Step S5 includes the following sub-steps: Step S501: Generate risk zoning early warning results based on the composite pollution risk index set; Step S502, decompose the contribution of the composite pollution risk index set based on the event propagation graph, specifically as follows: While keeping the event propagation weight set unchanged, a preset amplitude perturbation is applied to the potential release quantity index corresponding to each source node. Calculate the risk increments of algal bloom trigger probability, hypoxia risk probability, and secondary black and odorous environment risk probability before and after the disturbance; Define the risk increment as the contribution of the corresponding source strength node and sort them, then output the risk contribution sort set; Step S503: Based on the risk contribution ranking set and the preset resource constraint rules, the minimum intervention set is calculated.

5. The method for ecological and environmental pollution risk assessment based on big data as described in claim 4, characterized in that, The selection logic for the minimum intervention set is as follows: When the probability of algal bloom triggering or the probability of hypoxia risk are higher than the preset threshold, the interception measures corresponding to the source strength nodes ranked higher in the risk contribution ranking set are selected first, and the source control timing is matched with the window before the rainstorm. When the probability of secondary black and odorous pollution is higher than the preset threshold, oxygenation or scheduling measures corresponding to the receiving nodes with the highest risk contribution ranking will be selected first, and the repair sequence of the window after the rainstorm will be matched. The source control timing sequence and the repair timing sequence are combined and output as a minimum intervention set and input to the control terminal.

6. A big data-based ecological environment pollution risk assessment system, applied in any one of the big data-based ecological environment pollution risk assessment methods as described in claims 1-5, characterized in that, It includes a data acquisition module, a data inversion module, a propagation analysis module, a driving coupling module, and a risk assessment module; The data acquisition module is used to collect rainfall radar data, weather forecast data, remote sensing cover data, soil moisture data, topographic runoff data, agricultural activity proxy data, river and lake water quality data, and hydrological flow data. It then performs cleaning, time synchronization, and spatial gridding to form raw event perception data and construct an event-aligned dataset. The data inversion module is used to invert the potential release index set of nutrients in a sub-basin based on the event-aligned dataset. The propagation analysis module is used to construct an event propagation graph, calculate edge weights based on event alignment data, and output an event propagation weight set. The driving coupling module is used to input the potential release index set, the event propagation weight set, and the event alignment dataset into the event-driven coupling evaluation model, and output the transient inflow load prediction set and the composite pollution risk index set. The risk assessment module is used to generate risk zoning early warning results and risk contribution ranking sets based on the composite pollution risk indicator set, and select the corresponding minimum intervention set to output to the control end.

Images